Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!orstcs!usenet!jacobs.CS.ORST.EDU!keffert
From: keffert@jacobs.CS.ORST.EDU (Thomas Keffer)
Newsgroups: comp.lang.c++
Subject: Re: Why is this program slow? / C++ versus C Performance
Message-ID: <1991Jan12.225541.6709@usenet@scion.CS.ORST.EDU>
Date: 12 Jan 91 22:55:41 GMT
References: <1991Jan12.040453.6887@mentor.com>
Sender: @usenet@scion.CS.ORST.EDU
Distribution: usa
Organization: Oregon State University, CS Dept.
Lines: 42
Nntp-Posting-Host: jacobs.cs.orst.edu

In article <1991Jan12.040453.6887@mentor.com> timh@.mentorg.com (Tim Harvey @ APD x1381) writes:
>Peter Shirley posted the article (# 10640) "why is this program slow?"
>
>C++ is often charged with being an inherently slower language than C.
>
>A simple case was developed where a C++ program took almost twice as
>long (about 170%) to run as a similar C program.  It illustrates one
>situation where C++ appears slower than C.

Now hold on!  We have done extensive testing of our C++  Math.h++ class library
versus both C and Fortran and found only a very slight penalty
that can very easily be recovered by using the more sophisticated
algorithms made possible by the encapsulation, etc. properties of C++.

Here's some hard data.  The times shown are the times required to multiply
two matrices together of the given size.  The machine is a Sun 3/260
with a 68881 coprocessor.

N**2            g++ -m68881         f77 -f68881
25                  4.7s               3.9
100                16.4               15.5
400                64.2               62.1

You might want to implement a special "small matrix" class if you are
going to be working with very small N (2, 3 or 4) matrices (e.g., graphics
transformation matricies).  But for larger N (5 and up), the time
differences are completely trivial.

The trick is careful attention to detail: use reference counting, 
minimize memory allocations, maintain a pool of the small (reference)
structures, etc.

-tk

Disclaimer:
  I work for a company that vends a C++ math class.

___
Thomas Keffer
Rogue Wave
PO Box 2328 Corvallis, OR 97330
(503) 745-5908